EFRI BRAID: Brain-inspired Algorithms for Autonomous Robots (BAAR)

自主机器人，如无人驾驶汽车(SDVS)和家用协作机器人(CoBots)，具有巨大的潜力造福社会，满足国家的几个重要需求。尽管人工智能(AI)已经取得了实质性的进步，但与人类相比，自主机器人目前的数据/计算效率和适应性相形见绌，无法完成日常的感知运动任务，如驾驶和烹饪。使这种自主机器人能够不断地从经验中学习，并像人类一样在现实世界中持续提高它们的效率和弹性，这是它们广泛部署的当务之急。该项目旨在利用神经生物学学习和大脑智能的原理和见解，为机器人自主开发新的计算算法。研究结果可能会对自主机器人产生多方面和变革性的影响，如SDVS、CoBots和制造业和医疗保健应用中的其他智能机器人系统，这些机器人面临着计算/数据效率低下和适应不灵活的相同挑战。该项目寻求通过将大脑启发的智能融入到自主机器人系统的感知、规划和持续学习的基础和核心能力中，来提供自主机器人系统的范式转变。使用融合的工程科学方法，该项目旨在为自主机器人创建一个受大脑启发的感知、学习和规划算法的基本和创新框架。通过理论和实证相结合的研究，将该框架应用于SDVS和CoBots这两个具有代表性和互补性的工程系统。这项工作整合了受大脑启发的创新，将把一般受大脑启发的方法和算法改造和设计到SDVS和CoBot中，以进行数据和能效、适应性和弹性方面的有效性的实验验证。预计这些发现不仅将为SDVS和CoBot在现实世界中的部署提供重大飞跃，还将通过提高它们的数据/计算效率、适应能力和智能可解释性，对其他智能机器人系统(如制造和医疗保健领域的系统)产生革命性影响。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Autonomous robots, such as self-driving vehicles (SDVs) and household collaborative robots (Cobots), possess great potential to benefit society and meet several important national needs. Although artificial intelligence (AI) has made substantial progress, the current data/computational efficiency and adaptability of autonomous robots pale in comparison to humans in performing routine sensorimotor tasks such as driving and cooking. Enabling such autonomous robots to continually learn from experience and persistently improve their efficiency and resilience in the real world as humans do is imperative for their widespread deployments. This project aims to develop novel computational algorithms for robot autonomy with principles and insights of neurobiological learning and brain intelligence. The outcomes could make a multifaceted and transformative impact on autonomous robots such as SDVs, Cobots, and other intelligent robotic systems in manufacturing and healthcare applications that face the same challenges of computational/data inefficiency and adaptation inflexibility.The project seeks to provide a paradigm shift in autonomous robotic systems by incorporating brain-inspired intelligence throughout their fundamental and core capabilities of perception, planning, and continual learning. Using convergent engineering-science approaches, the project aims to create a fundamental and innovative framework of brain-inspired perception, learning, and planning algorithms for autonomous robots. The framework will be applied to SDVs and Cobots as two representative and complementary engineering systems through combined theoretical and empirical studies. Integrating brain-inspired innovations, the work will adapt and engineer the general brain-inspired methods and algorithms to SDVs and Cobots for experimental validation of the effectiveness in data- and energy-efficiency, adaptability, and resiliency. It is expected that the findings will not only provide a significant leap to SDVs and Cobots toward their real-world deployments, but also have a transformative impact on other intelligent robotic systems such as those in manufacturing and healthcare domains by improving their data/computation efficiency, adaptation resiliency, and intelligence interpretability.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.